In CASE of mystery

Erlend

Dear Forum, I have gone code-blind. What is wrong with below CASE section? - the compiler stops at 64 with "Expected and instruction or variable"

PUB Gain(gainSel) | bGain, regbyte                                                              

   CASE gainSel    
      128: bGain:= 7                                                          'gain selected - in dB
       64: bGain:= 6         
       32: bGain:= 5         
       16: bGain:= 4         
        8: bGain:= 3         
        4: bGain:= 2         
        2: bGain:= 1         
        1: bGain:= 0
        OTHER: RETURN FALSE          
    
   regByte:= i2c.ReadByteA8(ChipAddr, CTRL1_CRP) & !bGAINselCTRL1             'blank out the old Gain bits
   i2c.WriteByteA8(ChipAddr, PWR_CTRL, regByte | bGain)                       'write byte with new lower 3 bits representing Gain
   Recal                                                                      'always do re-cal after
   RETURN bGain

Erlend

LA6WNA

Not sure, but what about moving 64,32 and 16 one step left. To align the first digits with 1 in 128. The other numbers also is to be moved 2 step left.

Erlend

Ahh, I never fail to go into the indentation-trap! Thank you!

CASE gainSel    
      128: bGain:= 7                                                          'gain selected - in dB
      64: bGain:= 6         
      32: bGain:= 5         
      16: bGain:= 4         
      8: bGain:= 3         
      4: bGain:= 2         
      2: bGain:= 1         
      1: bGain:= 0
      OTHER: RETURN FALSE          
    
   regByte:= i2c.ReadByteA8(ChipAddr, CTRL1_CRP) & !bGAINselCTRL1             'blank out the old Gain bits
   i2c.WriteByteA8(ChipAddr, PWR_CTRL, regByte | bGain)                       'write byte with new lower 3 bits representing Gain
   Recal                                                                      'always do re-cal after
   RETURN bGain

Erlend

JonnyMac

The problem is assumed indenting levels. You can left-align the digits, or pad with zeroes to fix the alignment.

Since the outcome is a single variable update, I suggest you replace case with this line: (corrected)

  bGain := lookdownz(gainSel : 1, 2, 4, 8, 16, 32, 64, 128)

IF gainSel is in the list you'll get 0..7 back, otherwise you'll get 0 -- this is exactly what your case is doing.

Wuerfel_21

Assuming that you only actually pass in powers of two or zero, you can also do this:

bGain := ((>|gainSel)-1) #> 0

(other values will get rounded down to the nearest power of two)

JonnyMac

I had originally suggested what @Wuerfel_21 did -- this is better if you have control of gainSel and it's not going to be outside the list of values that you spell out in the case. What her suggestion does, if converted to case, is this:

  case gainSel
    000..001 : bGain := 0
    002..003 : bGain := 1
    004..007 : bGain := 2
    008..015 : bGain := 3    
    016..031 : bGain := 4    
    032..063 : bGain := 5   
    064..127 : bGain := 6    
    128      : bGain := 7    
    other    : bGain := 0

That said, there is no upper limit, so you may want to modify like this:

  bGain := 0 #> >|gainSel-1 <# 7

Warning: If gainSel goes negative, you're always going to get 7 for bGain. You could fix that with:

  if (gainSel > 0)
    bGain := 0 #> >|gainSel-1 <# 7
  else
    bGain := 0

Erlend

JonnyMac wrote: »

The problem is assumed indenting levels. You can left-align the digits, or pad with zeroes to fix the alignment.

Since the outcome is a single variable update, I suggest you replace case with this line:

  bGain := lookupz(gainSel : 1, 2, 4, 8, 16, 32, 64, 128)

IF gainSel is in the list you'll get 0..7 back, otherwise you'll get 0 -- this is exactly what your case is doing.

I like this solution a lot. I have several of these parameter list situations, and not all of them are regular like this one.

Erlend

Erlend

I did this, though - to not write to the register if not valid

   bGain := LOOKUP(gainSel : 1, 2, 4, 8, 16, 32, 64, 128)                     'convert from desired gain dB to bit numbers to set
   IF bGain == 0                                                              'if not found, do not write to register
     RETURN FALSE
   bGain -= 1                                                                 'because I used lookup instead of lookupz
   regByte:= i2c.ReadByteA8(ChipAddr, CTRL1_CRP) & !bGAINselCTRL1             'blank out the old Gain bits
   i2c.WriteByteA8(ChipAddr, PWR_CTRL, regByte | bGain)                       'write byte with new lower 3 bits representing Gain
   Recal                                                                      'always do re-cal after
   RETURN bGain

Agree?

Wuerfel_21

Actually, doesn't it have to be LOOKDOWN instead of LOOKUP?

JonnyMac

Actually, doesn't it have to be LOOKDOWN instead of LOOKUP?

You're absolutely right. Clearly, the coffee hadn't kicked in when I made that suggestion this morning. Thanks for the catch.

Erlend

JonnyMac wrote: »

Actually, doesn't it have to be LOOKDOWN instead of LOOKUP?

You're absolutely right. Clearly, the coffee hadn't kicked in when I made that suggestion this morning. Thanks for the catch.

Now I have read the manual (again) and gotten to know that LOOKUP(Z) takes an index and returns the corresponding number in the list, whereas LOOKDOWN(Z) takes a number and returns its index in the list. I need the latter. Thanks @JonnyMac and @Wuerfel_21

   bGain := LOOKDOWN(gainSel : 1, 2, 4, 8, 16, 32, 64, 128)                     'convert from desired gain dB to bit numbers to set
   IF bGain == 0                                                              'if not found, do not write to register
     RETURN FALSE
   bGain -= 1                                                                 'because I used lookdown instead of lookdownz
   regByte:= i2c.ReadByteA8(ChipAddr, CTRL1_CRP) & !bGAINselCTRL1             'blank out the old Gain bits
   i2c.WriteByteA8(ChipAddr, CTRL1_CRP, regByte | bGain)                       'write byte with new lower 3 bits representing Gain
   Recal                                                                      'always do re-cal after
   RETURN gainSel

Erlend

For the lists that are not contiguous i settled for this code, 'padding' the list with -1, - but it is maybe not very elegant.

bSPS := LOOKDOWN(SPSsel : 10, 20, 40, 80, -1, -1, -1, 320)
IF bSPS == 0                                              
  RETURN FALSE                                            
bSPS -= 1

Erlend

JonnyMac

If spsSel can never be -1, and the only valid values for bSPS are 0, 1, 2, 3, and 7, that makes sense to me.